Online Sales Forecasting with the Multiple Regression Analysis Data Matrices Package

Abstract

The Web-based Multiple Regression Analysis Data Matrices Package (developed jointly with Justin Yost) enables competing participant teams in the marketing simulation COMPETE to apply their knowledge of multiple regression analysis in sales forecasting. Participants with Web-access use this package to create nine data matrices (one data matrix for each strategic business unit) consisting of relevant predictor and response variables for each of the prior decision periods. Next, the data are screened for potential multicollinearity using correlation analysis. Then, the top two predictor variables that satisfy multiple regression analysis assumptions are used to build a linear unrestricted single-equation multiple regression model. The results are checked for potential heteroskedasticity.